model_stage2 / README.md
huudan123's picture
Add new SentenceTransformer model.
78216f6 verified
metadata
base_model: huudan123/model_stage1
datasets: []
language: []
library_name: sentence-transformers
metrics:
  - pearson_cosine
  - spearman_cosine
  - pearson_manhattan
  - spearman_manhattan
  - pearson_euclidean
  - spearman_euclidean
  - pearson_dot
  - spearman_dot
  - pearson_max
  - spearman_max
pipeline_tag: sentence-similarity
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:183796
  - loss:MultipleNegativesRankingLoss
widget:
  - source_sentence: >-
      nếu thời_gian đến mà họ phải có một cuộc đấu_tranh johny shanon có_thể là
      một người ngạc_nhiên
    sentences:
      - johny nghĩ anh ta  người giỏi nhất trong thị_trấn
      - nếu một cuộc đấu_tranh đã xảy ra johny có_thể ngạc_nhiên đấy
      - tất_cả bằng_chứng về văn_hóa từ xã_hội của umbria đã bị mất
  - source_sentence: chèn jay leno đùa  đây
    sentences:
      - mathews đã chỉ ra rằng sẽ không cần phải tuyển_dụng luật_sư địa_phương
      - đây  nơi  một trò_đùa jay leno sẽ đi
      - jay leno không phải  một diễn_viên hài
  - source_sentence: đúng_vậy tất_cả  lỗi của họ
    sentences:
      - bạn bị giới_hạn bởi số_lượng bộ_nhớ bạn đã 
      - phải tất_cả đều  lỗi của họ
      - rõ_ràng  tất_cả những lỗi của công_nhân
  - source_sentence: >-
      6 mặc_dù mỗi cơ_quan phát_triển và triển_khai các thỏa_thuận hiệu_quả
      phản_ánh các ưu_tiên tổ_chức cụ_thể cấu_trúc và nền văn_hóa các thỏa_thuận
      hiệu_quả đã gặp các đặc_điểm sau
    sentences:
      - các thỏa_thuận hiệu_quả đã được phát_hành từ mỗi đại_lý
      - kế_hoạch hiệu_quả loại_trừ bất_cứ điều  để làm với các cấu_trúc
      - không   bên trong sảnh trên đồi cả
  - source_sentence: hay na uy hay  đó
    sentences:
      - na uy hay cái  đó khác
      - điều đó hoàn_toàn không đúng
      - na uy hoặc từ một trong những quốc_gia scandinavia
model-index:
  - name: SentenceTransformer based on huudan123/model_stage1
    results:
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: sts evaluator
          type: sts-evaluator
        metrics:
          - type: pearson_cosine
            value: 0.6279986884327646
            name: Pearson Cosine
          - type: spearman_cosine
            value: 0.6257861952118347
            name: Spearman Cosine
          - type: pearson_manhattan
            value: 0.6286844662908612
            name: Pearson Manhattan
          - type: spearman_manhattan
            value: 0.6309663003206769
            name: Spearman Manhattan
          - type: pearson_euclidean
            value: 0.6277475064516767
            name: Pearson Euclidean
          - type: spearman_euclidean
            value: 0.6297451268540156
            name: Spearman Euclidean
          - type: pearson_dot
            value: 0.588316765453479
            name: Pearson Dot
          - type: spearman_dot
            value: 0.5802157556789215
            name: Spearman Dot
          - type: pearson_max
            value: 0.6286844662908612
            name: Pearson Max
          - type: spearman_max
            value: 0.6309663003206769
            name: Spearman Max

SentenceTransformer based on huudan123/model_stage1

This is a sentence-transformers model finetuned from huudan123/model_stage1. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

  • Model Type: Sentence Transformer
  • Base model: huudan123/model_stage1
  • Maximum Sequence Length: 512 tokens
  • Output Dimensionality: 768 tokens
  • Similarity Function: Cosine Similarity

Model Sources

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: RobertaModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("huudan123/model_stage2")
# Run inference
sentences = [
    'hay na uy hay gì đó',
    'na uy hoặc từ một trong những quốc_gia scandinavia',
    'na uy hay cái gì đó khác',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Semantic Similarity

Metric Value
pearson_cosine 0.628
spearman_cosine 0.6258
pearson_manhattan 0.6287
spearman_manhattan 0.631
pearson_euclidean 0.6277
spearman_euclidean 0.6297
pearson_dot 0.5883
spearman_dot 0.5802
pearson_max 0.6287
spearman_max 0.631

Training Details

Training Hyperparameters

Non-Default Hyperparameters

  • overwrite_output_dir: True
  • eval_strategy: epoch
  • per_device_train_batch_size: 256
  • per_device_eval_batch_size: 256
  • num_train_epochs: 15
  • warmup_ratio: 0.1
  • fp16: True
  • load_best_model_at_end: True
  • gradient_checkpointing: True

All Hyperparameters

Click to expand
  • overwrite_output_dir: True
  • do_predict: False
  • eval_strategy: epoch
  • prediction_loss_only: True
  • per_device_train_batch_size: 256
  • per_device_eval_batch_size: 256
  • per_gpu_train_batch_size: None
  • per_gpu_eval_batch_size: None
  • gradient_accumulation_steps: 1
  • eval_accumulation_steps: None
  • learning_rate: 5e-05
  • weight_decay: 0.0
  • adam_beta1: 0.9
  • adam_beta2: 0.999
  • adam_epsilon: 1e-08
  • max_grad_norm: 1.0
  • num_train_epochs: 15
  • max_steps: -1
  • lr_scheduler_type: linear
  • lr_scheduler_kwargs: {}
  • warmup_ratio: 0.1
  • warmup_steps: 0
  • log_level: passive
  • log_level_replica: warning
  • log_on_each_node: True
  • logging_nan_inf_filter: True
  • save_safetensors: True
  • save_on_each_node: False
  • save_only_model: False
  • restore_callback_states_from_checkpoint: False
  • no_cuda: False
  • use_cpu: False
  • use_mps_device: False
  • seed: 42
  • data_seed: None
  • jit_mode_eval: False
  • use_ipex: False
  • bf16: False
  • fp16: True
  • fp16_opt_level: O1
  • half_precision_backend: auto
  • bf16_full_eval: False
  • fp16_full_eval: False
  • tf32: None
  • local_rank: 0
  • ddp_backend: None
  • tpu_num_cores: None
  • tpu_metrics_debug: False
  • debug: []
  • dataloader_drop_last: False
  • dataloader_num_workers: 0
  • dataloader_prefetch_factor: None
  • past_index: -1
  • disable_tqdm: False
  • remove_unused_columns: True
  • label_names: None
  • load_best_model_at_end: True
  • ignore_data_skip: False
  • fsdp: []
  • fsdp_min_num_params: 0
  • fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
  • fsdp_transformer_layer_cls_to_wrap: None
  • accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
  • deepspeed: None
  • label_smoothing_factor: 0.0
  • optim: adamw_torch
  • optim_args: None
  • adafactor: False
  • group_by_length: False
  • length_column_name: length
  • ddp_find_unused_parameters: None
  • ddp_bucket_cap_mb: None
  • ddp_broadcast_buffers: False
  • dataloader_pin_memory: True
  • dataloader_persistent_workers: False
  • skip_memory_metrics: True
  • use_legacy_prediction_loop: False
  • push_to_hub: False
  • resume_from_checkpoint: None
  • hub_model_id: None
  • hub_strategy: every_save
  • hub_private_repo: False
  • hub_always_push: False
  • gradient_checkpointing: True
  • gradient_checkpointing_kwargs: None
  • include_inputs_for_metrics: False
  • eval_do_concat_batches: True
  • fp16_backend: auto
  • push_to_hub_model_id: None
  • push_to_hub_organization: None
  • mp_parameters:
  • auto_find_batch_size: False
  • full_determinism: False
  • torchdynamo: None
  • ray_scope: last
  • ddp_timeout: 1800
  • torch_compile: False
  • torch_compile_backend: None
  • torch_compile_mode: None
  • dispatch_batches: None
  • split_batches: None
  • include_tokens_per_second: False
  • include_num_input_tokens_seen: False
  • neftune_noise_alpha: None
  • optim_target_modules: None
  • batch_eval_metrics: False
  • eval_on_start: False
  • batch_sampler: batch_sampler
  • multi_dataset_batch_sampler: proportional

Training Logs

Epoch Step Training Loss loss sts-evaluator_spearman_max
0 0 - - 0.6283
0.6964 500 4.3237 - -
1.0 718 - 2.3703 0.6500
1.3928 1000 2.2259 - -
2.0 1436 - 2.2597 0.624
2.0891 1500 2.0143 - -
2.7855 2000 1.7433 - -
3.0 2154 - 2.3027 0.6405
3.4819 2500 1.5279 - -
4.0 2872 - 2.3583 0.6094
4.1783 3000 1.3796 - -
4.8747 3500 1.2096 - -
5.0 3590 - 2.4877 0.6069
5.5710 4000 1.036 - -
6.0 4308 - 2.5685 0.6310
  • The bold row denotes the saved checkpoint.

Framework Versions

  • Python: 3.10.12
  • Sentence Transformers: 3.0.1
  • Transformers: 4.42.4
  • PyTorch: 2.3.1+cu121
  • Accelerate: 0.33.0
  • Datasets: 2.20.0
  • Tokenizers: 0.19.1

Citation

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}

MultipleNegativesRankingLoss

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply}, 
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}